RESEARCH ARTICLE

Build it and they will come, but not all of them

in fragmented Atlantic Forest landscapes
Marcio S. Suganuma1,2,3 , Giselda Durigan4,5

Ecological restoration interventions, in most cases, aim to restore habitat structure and plant community composition, thus
re-establishing ecosystem functioning as similar as possible to that of the pre-existing natural ecosystem. However, given the
difficulty of cultivating many species, that goal seems unattainable, unless “if you build it, they will come.” Here we addressed
the Field of Dreams hypothesis in the context of trees and shrubs of the seasonal tropical forest in southeastern Brazil. We aimed
to verify if the species from the regional pool have been able to colonize the restoration forests and if functional patterns exist
behind successful and unsuccessful colonization. We categorized each species by dispersal syndrome, seed traits, growth rate,
shade tolerance, and rarity in the communities. Most, but not all, species from the regional pool are colonizing forest patches
undergoing restoration. Successful colonizers are mostly zoochorous, dispersed by birds or bats, shade tolerant, of moderate
or fast growth. By partially confirming the Field of Dreams hypothesis, our study implies that it is not necessary to reintroduce
a large proportion of the regional pool in tropical forest restoration projects, given that many species will later spontaneously
arrive, even in fragmented landscapes. However, the existence of a particular functional profile (slow-growing species, dis-
persed by gravity or large mammals) that will rarely colonize the restoration forests should not be disregarded. Even though
these are a minority, such species will likely be confined to the remaining natural fragments if they are not included in the res-
toration projects.
Key words: assembly rules, ecological filters, Field of Dreams hypothesis, forest restoration, resilience, understory colonization

into the practice of restoration. Among those hopes is The Field

Implications for Practice
of Dreams hypothesis (Palmer et al. 1997). This hypothesis
• Most species occurring in native remnants of tropical for- assumes that once habitat structure has recovered, biodiversity
ests are able to colonize patches of neighboring forests will spontaneously return. It was included among the myths of
undergoing restoration. Whether we plant them or not, restoration (Hilderbrand et al. 2005), as it assumes a certain pre-
they will likely arrive later. dictability in communities’ trajectories, ignoring uncertainty.
• Trees and shrubs dispersed by birds or bats, shade toler- However, more than 20 years later, the hypothesis has been
ant, and do not grow slowly are the most successful
tested for few organisms and few ecosystems, with contrasting
colonizers.
results, not allowing generalizations to be made (e.g., Sudduth
• Species dispersed by gravity or terrestrial mammals, slow
et al. 2011 for aquatic ecosystems, Frick et al. 2014 for pollina-
growing, or naturally rare have a lower chance to colonize
tors, and Wodika & Baer 2015 for soil macrofauna). To our
the restoration forests, requiring special assistance to per-
knowledge, the Field of Dreams hypothesis was never properly
sist in the landscape.
tested for plant species in tropical forest restoration, despite
• The dimensions of the diaspore do not accurately predict
recent studies showing that tree biodiversity tends to increase
if a tree species will or will not successfully colonize for-
est islands undergoing restoration.
Author contributions: GD conceived the research; MSS, GD designed the research and
collected the data; MSS organized and analyzed the data; MSS, GD interpreted the
results, wrote, and edited the manuscript.
1
Laborat
orio de Biodiversidade e Restauraç~ao de Ecossistemas, Universidade Estadual
Introduction de Londrina, Londrina, Brazil
2
Centro de Ciências Humanas e da Educaç~ao, Universidade Estadual do Norte do
The expectation that restoration actions reestablish all species Paran
a, Cornélio Proc
opio, Brazil
3
from the pre-existing ecosystem has never been abandoned, Address correspondence to M. S. Suganuma, email marciosuganuma@gmail.com
4
Laborat
orio de Ecologia e Hidrologia Florestal, Floresta Estadual de Assis, Instituto
even though research has demonstrated the constraints posed Florestal, Assis, Brazil
by ecological and socioeconomic filters limiting the range of
5
Instituto de Biologia, Universidade Estadual de Campinas (UNICAMP), Campinas,
Brazil
species whose reintroduction is possible in a given situation
(Rey-Benayas et al. 2009; Suding 2011; Maron et al. 2012; © 2021 Society for Ecological Restoration.
doi: 10.1111/rec.13537
Ren et al. 2017; Jones et al. 2018). However, recent studies have Supporting information at:
also brought good news and paradigm shifts, injecting optimism http://onlinelibrary.wiley.com/doi/10.1111/rec.13537/suppinfo

Restoration Ecology 1 of 10
Build it and most will come

over time across tropical forest restoration (Suganuma & Here we addressed the Field of Dreams hypothesis in the con-
Durigan 2015; Crouzeilles et al. 2016; Li et al. 2018) and that text of trees and shrubs of the seasonal tropical forest (SFT) in
the earlier forest structure is recovered, the earlier biodiversity southeastern Brazil. We aimed to assess if the species recorded
recovery will begin (Wheeler et al. 2016; Brancalion et al. 2019). in the native forest remnants have been able to colonize nearby
Among the recent paradigm shifts in restoration ecology, one restored forests without human assistance. We categorized each
widespread trend is to look at ecosystems from the perspective species by dispersal syndrome, seed traits, growth rate, shade
of functional traits of organisms instead of looking at taxonomic tolerance, and rarity in the communities and analyzed guild pro-
diversity (Laughlin 2014; Carlucci et al. 2020). This novel per- portionality in search for functional patterns behind successful
spective is advantageous when used to identify key functions and unsuccessful colonization. We hypothesize that most spe-
behind community assembly mechanisms and apply these func- cies from the regional pool are able to colonize forest patches
tions as tools to reduce costs, increase probability of successful undergoing restoration, but some species are not able to cross
restoration, and optimize ecosystem services. Applying this the matrix or do not establish, staying confined to natural frag-
approach, however, is limited by gaps in our understanding of ments or to those sites where they were planted (H1) and that
ecosystem functioning. There are many unknown functions ecological filters lead to the convergence of attributes among spe-
of individual species, which will be neglected if species are con- cies that successfully colonize forests undergoing restoration, by
sidered redundant based on a few traits, because redundancy in barring those with unsuitable functional profiles (H2).
one function does not imply redundancy in any other function
(Naeem 1998). If a greater number of traits is considered, spe-
cies richness and diversity can be strong predictors of functional Methods
diversity (Lohbeck et al. 2012). There is also a risk of seeking
complementarity among attributes that are irrelevant to ecosys- Study Region
tem functioning in the particular conditions to be restored, given This study was developed from an extensive dataset of occur-
the incomplete and sometimes superficial understanding of the rence and abundance of tree and shrub species in nine native
effect or response of a given attribute to different factors remnants of STF and 26 STF patches being restored. The sites
(Lohbeck et al. 2012). are located in the states of S~ao Paulo, Paran
a and Mato Grosso
The Field of Dreams hypothesis relies upon the ability of do Sul, Brazil, across an area of approximately 80,000 km2
native species to colonize the areas undergoing restoration, (Fig. 1, each site described in detail in Table S1). In this region,
spontaneously recovering biodiversity. Colonization results native remnants are restricted to about 10% of their original
from a balance between the characteristics of the species dia- range, and the average distance between remnants is nearly
spores (which allow them to arrive) and how the species respond 1,000 m (Nalon et al. 2020). The same dataset was previously
to habitat conditions (allowing establishment) (Duarte explored by Suganuma and Durigan (2015) for modeling the
et al. 2011; Myers & Harms 2011). Which species will success- successional trajectories of the restoration forests, and later by
fully colonize the restoration sites depends, ultimately, on biotic Suganuma et al. (2017) in search for the drivers of restoration
and abiotic constraints acting locally and the functional traits of success. The STF is one of the sub-types of the Brazilian Atlan-
the species that allow them to overcome those filters (Palmer tic Forest, characterized by a seasonal climate, with rainfall con-
et al. 1997; Lebrija-Trejos et al. 2010; Burton et al. 2011). The centrated in the summer and a dry season in the winter. The three
importance of fauna in restoration of tropical forests has been Köppen climate types in the region—Cfa, Cwa, and Aw—differ
globally discussed and the key role of dispersal agents in driving especially in the duration of the dry season and mean annual
succession in restoration plantings is widely known temperatures, which increase slightly from the first to the last cli-
(Catterall 2018; Cross et al. 2019). However, seed dispersers mate type. The mean annual temperature in the study sites
vary in their mobility and functions, so the effective restoration ranges from 21 to 24 C, with minimums around 10 C and max-
agents are often a subset of the whole fauna existing in the imums around 30 C. The average annual rainfall varies between
region (Beltr
an & Howe 2020). The degree of isolation from 1,150 and 1,630 mm, concentrated in the summer, and can be
seed sources, absence of animal seed dispersers, and large seed less than 40 mm in some winter months (Sentelhas
size are all limitations to seed dispersal to restoration sites et al. 2003). Despite all sample areas being in riparian zones,
(Reid et al. 2015; Aavik & Helm 2018; Piotto et al. 2019). In from 0 to 50 m away from river margins, they were all in well
addition, there is a trade-off between favorable seed traits for drained terrain free from floods.
successful dispersal versus successful establishment, because The forests undergoing restoration ranged from 4 to 53 years
large seed size is advantageous for seed germination and seed- old (Table S1) and replaced land use mostly as coffee plantation,
ling establishment within tropical forests (Moles et al. 2007). crops, or sugar cane for several decades (sometimes centuries).
Luckily, recent work has provided some good news: the distance No contribution, therefore, could be expected from the seed
threshold for native remnants to serve as seed sources for neigh- bank or from underground structures resprouting from a pre-
boring restoration sites is much larger than initially thought existing forest. Sites differed in soil types (texture and fertility)
(Kirmer et al. 2008; Suganuma et al. 2017). Therefore, many as well as different pre-restoration land use (pasture, sugarcane,
species of fauna and flora can successfully move between or crops). Nursery-raised seedlings of multiple tree species were
islands in the landscape even in severely fragmented landscapes planted in the restoration sites with varying density, richness,
(Gonz
alez-Varo et al. 2017). species composition, management practices, and distance from

2 of 10 Restoration Ecology
 Build it and most will come

Figure 1. Location of the 35 study sites in the states of S~ao Paulo, Paran
a and Mato Grosso do Sul, Brazil (adapted from Suganuma & Durigan 2015).

seed sources. The distance between each restoration site to the of three groups of species: (1) Successful colonizers (SC)—
nearest forest remnant was on average 1,174 m, ranging from species spontaneously regenerating in at least one site where that
0 to 5,700 m. The nine remnants of native forests sampled rep- species was not planted. Such species overcame filters related to
resent the distinct environmental conditions (soil and climate) both seed dispersal and seedling establishment; (2) non-
and level of past disturbances within the study region. Although colonizers (NC)—species occurring in at least one of the native
some rare species occurring in the region likely escaped our forests, which were not recorded in any of the restoration sites.
sampling, we assumed that the huge number of 15,306 individ- Species forming this group were not able to cross the matrix
uals and 196 species sampled in these nine widely distributed (seed dispersal limitation) or could not germinate or survive in
remnants provided a good representation of the functional guilds the restoration forests; and (3) regional species pool (RP)—all
of the native forests in the study region. native species sampled in the nine patches of native forests or
in the 26 patches of restoration forests (planted or spontaneously
regenerating). All these species could, potentially, colonize the
Data Collection
restoration forests where they were not planted.

At each study site, we sampled a total area of 1,000 m2, subdi-

vided into 10 plots of 20  5 m, randomly distributed in a strip
50 m wide along the river margins, stretching at least 200 m and Functional Guilds
no more than 700 m. We sampled, in total, 9,000 m2 of native We searched for species traits or functions related to the ecolog-
forests and 26,000 m2 of restoration forests. We identified and ical filters, which constrain seed dispersal or seedling establish-
counted all woody plants (trees and shrubs) with a minimum ment. Functional guilds explored are related to adaptations of
height of 0.50 m in the whole sampled area. According to the species to the environment, and taken together they reveal
planting row alignment and the list of species planted in each ecological differentiation between species (Götzenberger
restoration project, we categorized each sampled individual as et al. 2012). We categorized each species by dispersal syndrome
“planted” or “regenerating” and identified them to species level. as zoochorous, anemochorous, or autochorous based on litera-
Exotic species (not occurring naturally in the STF) were ture (Van der Pijl 1972; Almeida-Neto et al. 2008) (Table S2).
excluded from this study. Since we aimed to identify colonizers The animal-dispersed species were later categorized in four sub-
crossing the matrix, we also eliminated from the functional ana- sets, according to the main disperser: birds, bats, terrestrial
lyses all individuals regenerating within a restoration forest mammals, and mixed (more than one syndrome combined)
patch if that species was planted in that particular site, since (Van der Pijl 1972; Almeida-Neto et al. 2008). Growth rate (fast,
these individuals could have originated from the planted trees moderate, or slow) and shade tolerance (tolerant or intolerant)
and not from a colonization process. are related to ecological filters driving seedling establishment
To verify if there is a functional bias in the colonization pro- and recruitment, and species were categorized in these guilds
cess, we analyzed the frequency of functional guilds within each based upon literature (Carvalho 2003, 2006, 2008; Durigan

Restoration Ecology 3 of 10
Build it and most will come

Figure 2. Proportion of species (left) and individuals (right) classified according to dispersal syndrome and by subset of zoochory type, in three groups of species:
Regional pool (all native species recorded, except those planted), successful colonizers (species that successfully colonized restoration sites), and non-colonizers
(species from the regional pool not colonizing restoration sites), in a set of 26 restoration sites and nine remnants of seasonal tropical forest, southeastern Brazil.
Dispersal syndromes among species (A), dispersal syndromes among individuals (B), subsets of zoochory syndrome among species (C), and subsets of zoochory
syndrome among individuals (D). The number on top of each column is the frequency in percentage. Green asterisk (*) means the frequency of that guild
significantly differs from the frequency in the regional pool (α = 0.05).

et al. 2004) (Table S2). We also described species traits likely correction for n < 5 and Fisher exact test (α = 0.05). We inves-
related to both seed dispersal and seedling establishment, tigated if the frequency of each functional guild within the group
including: seed mass (g), length (mm), and surface area (mm2) of species successfully colonizing the restoration patches
of the diaspore. Seed traits were obtained primarily from litera- (SC) or in the group of species never recorded as colonizers
ture (Lorenzi 1998, 2002, 2009; Carvalho 2003, 2006, 2008; (NC), differed from the frequency of that guild in the regional
Almeida-Neto et al. 2008; Mori et al. 2012) and supported by species pool (RP). We compared the frequency of each func-
on line data bases (IPEF 2016; Species Link 2016). Addition- tional guild among species (relative contribution in richness)
ally, we applied the rarity criteria proposed by Caiafa and Mar- and among individuals (relative contribution in abundance).
tins (2010), which assess local abundance, habitat affinity, and We carried out the two analyses assuming that abundance data
geographic distribution, to categorize the species as rare or com- implicitly represent fitness and dominance of each guild within
mon, since it could influence the abundance of diaspores in the the communities, while niche complementarity is more
landscape (diaspore pressure) and, therefore, the probability of related to presence/absence of species representing different
colonization. functional guilds. We assumed that if the frequency of a func-
tional guild within one group (SC or NC) does not differ from
the regional pool that indicates no association between that
Data Analyses trait or function and the ecological filters hindering dispersal
We used chi-square analysis (α = 0.05) to verify if colonization or establishment.
success is associated with functional profile of the species (cate- We assessed the relationships between diaspore traits and
gorical variables), following a guild-proportionality approach, as colonization success by logistic regression for each trait sepa-
recommended by Götzenberger et al. (2012). We applied Yates rately. The binary response variable was colonizing success

4 of 10 Restoration Ecology
 Build it and most will come

Figure 3. Proportion of species (left) and individuals (right) classified according to functional traits and rarity in three groups of species: Regional pool (all native
species recorded, except those planted), successful colonizers (species that successfully colonized restoration sites), and non-colonizers (species from the regional
pool not colonizing restoration sites), in a set of 26 restoration sites and nine remnants of seasonal tropical forest, southeastern Brazil. Growth rate among species
(A), growth rate among individuals (B), shade tolerance among species (C), shade tolerance among individuals (D), native range among species (E), and native
range among individuals (F). The number on top of each column is the frequency in percentage. Green asterisk (*) means the frequency of that guild significantly
differs from the frequency in the regional pool (α = 0.05).

(1) or failure (0). Diaspore length (mm), diaspore surface area regressions separately for each group of plant species according
(mm2), and seed mass (g), for each species, were individually to dispersal type for anemochory and zoochory (birds and bats).
explored as predictor variables. Based on the premise that the To verify the robustness of each relationship, we applied the
dimensions of the diaspore and the seed mass can have influence likelihood ratio test (α = 0.05). All analyses were performed
between groups of dispersal agents, we performed also logistic in R (version 4.0.1) with packages mass and vegan.

Restoration Ecology 5 of 10
Build it and most will come

Results ( 21% species and 88% individuals), shade-intolerant

The regional species pool (RP) was represented by 196 ( 43% species and 75% individuals), and common species
native tree and shrub species (15,306 individuals sampled) (the ( 11% individuals).
complete species list in Table S2). The regional community In general, a guild that was more frequent among SC than in
assemblages were dominated by shade tolerant (79% of species the RP was less frequent among NC than in the RP and vice
and 88% of individuals) and zoochorous species (75% of spe- versa. However, two guilds—anemochorous and shade
cies and 68% of individuals). A subset of 43 species (22%) from intolerant—were sampled in significantly lower frequencies
the regional pool (3,360 individuals sampled) occurred only in than in the RP either among SC or NC, because they are rela-
the forest remnants and did not colonize the restoration sites tively more abundant in the overstory than among young plants
(NC). A sub-set of 79 species (40%) from the regional pool regenerating.
(6,579 individuals sampled) had colonized restoration sites where We found no significant relationships when we used logistic
they were not planted (SC). The remaining 74 species (38%), regressions to explore the colonization probability for all species
despite being observed spontaneously regenerating in the planted in the regional pool as a function of seed mass, seed area, and
forests, were not categorized as successful or unsuccessful colo- diaspore length among anemochory species and zoochory spe-
nizers. These species were sampled in the native forests, but they cies (birds and bats), additional information in Table S4.
were also planted, making it impossible to know if regenerating
individuals were descendants from planted trees or had arrived
from nearby forests remnants. Discussion
Zoochory (especially by bats) was more frequent among col- The Field of Dreams hypothesis permeates, in some way, all
onizers than in the regional pool, while autochory and dispersed ecological restoration projects, because, if confirmed, the results
by terrestrial mammals were more frequent among NC than in of interventions aimed at the recovery of biodiversity would go
the RP. Colonizers were also more likely to be fast growing far beyond the direct effort of reintroduction undertaken in the
and shade tolerant. Differences in guild proportionality were projects. However, studies already carried out to test that
less pronounced when presence/absence of the species hypothesis have produced contrasting results for different eco-
(Figs. 2 & 3 left, Table S3) was considered instead of abun- systems and different organisms, preventing generalization.
dance (Figs. 2 & 3 right, Table S3). When the group of SC For instance, the return of biodiversity after restoring vegetation
was analyzed by dispersal guild proportionality (Fig. 2, structure has been demonstrated for soil arthropod (Meloni &
Table S3) compared to the regional pool, we found higher fre- Varanda 2015) and soil macrofauna in tropical forest
quency for zoochory (+13% species and +37% individuals). (Amazonas et al. 2018), herpetofauna in longleaf pine
Among sub-groups of dispersers, differences were found only forest (Litt et al. 2001), while, in contrast, neither amphibians
for bats, with +83% species and +145% individuals, compared in tropical montane forests (Díaz-García et al. 2017), inverte-
to the regional pool. SC had fewer representatives than the brates in boreal mires (Noreika et al. 2015), nectar-feeding birds
regional pool of the following dispersal guilds: anemochorous in Banksia woodlands (Frick et al. 2014), soil microbial fauna of
( 62% individuals) and authocorous ( 78% species and bottomland hardwood forests (Strickland et al. 2017), nor soil
89% individuals), and those dispersed by terrestrial mammals invertebrates of tallgrass prairie (Wodika & Baer 2015)
( 78% individuals). For growth and shade tolerance guilds responded to the recovery of vegetation structure. In this study,
(Fig. 3, Table S3), we found higher frequency among SC than we aimed to verify whether, in the severely fragmented land-
in the regional pool for fast- (+62% individuals) and moder- scape of the Atlantic Forest, the reconstruction of the forest
ate-growing (+16% species and +18% individuals), and structure through the planting of a subset of regional species
shade-tolerant plants (+4% species and +10% individuals). would trigger the arrival of new species of non-planted trees
Some other guilds were less frequent among SC than in the and shrubs, naturally expanding biodiversity. Previous studies
RP: slow growing ( 10% species and 26% individuals) and showing that tree richness of tropical forests undergoing restora-
shade intolerant ( 14% species and 67% individuals). tion tends to increase over time (Suganuma & Durigan 2015;
Among NC, the following dispersal guilds were more fre- Crouzeilles et al. 2016; Li et al. 2017) allowed us to expect a suc-
quent than in the RP (Fig. 2, Table S3): autochory (+111% spe- cessful colonization process.
cies and +395% individuals) and dispersal by terrestrial Our results partially confirmed the Field of Dreams hypothe-
mammals (+256%). Dispersal guilds, which were less frequent sis, even in the severely fragmented landscape of the Atlantic
among NC than in the RP, were: zoochory ( 13% species and Forest, where the average distance among forest remnants is
66% individuals), anemochory ( 84% individuals), and dis- about 1,000 m (Nalon et al. 2020). Most species from the
persal by bats ( 82% individuals). Higher frequency among regional pool were recorded in natural regeneration under the
NC than in the regional pool was recorded also for some guilds restoration forests, confirming H1. Natural regeneration pro-
not related with dispersal (Fig. 3, Table S3): slow-growing cesses within restoration forests have been impaired by the slow
(+19% species and +121% individuals), shade-tolerant arrival of diaspores (Reid et al. 2014; Aavik & Helm 2018;
(+11% species and +10% individuals), and rare species Piotto et al. 2019), which is influenced by distance from seed
(+32% individuals). Frequency was lower among NC than in sources, but successful colonization depends also on overcom-
the RP for fast- ( 57% individuals) and moderate-growing ing ecological filters related to seed germination and

6 of 10 Restoration Ecology
 Build it and most will come

competition (Lebrija-Trejos et al. 2010; Burton et al. 2011). For to expect that slow-growing individuals require more time than
this reason, Reid and Holl (2013) emphasized the importance of the age of the restoration forests studied here (4–53 years) to
long-term research to demonstrate the establishment of seed- reach a similar relative abundance as that found in the regional
lings in restoration projects, rather than solely the arrival of species pool.
diaspores. Rarity of a species in natural ecosystems did not appear to be
Distinct and meaningful functional profiles were found a relevant issue in the assembly of the restoration forests studied,
among successful and unsuccessful colonizers (confirming since the frequency of rare or common species or individuals did
H2), related to dispersal and establishment of plant species. SC not differ between SC and the regional pool. Except for relative
are mostly animal-dispersed (birds and bats), with dispersal by abundance slightly higher for rare and lower for common spe-
bats particularly increasing the chance of a species to colonize. cies among unsuccessful colonizers than in the regional pool,
Colonization has been favored also by being shade tolerant our results indicate that rarity is an emergent property of plant
and not growing slowly, both advantages to establish in the populations within the communities of forest islands in this
competitive environment of these restoration forests (Oliveira region, no matter if they are native or restored. Maina and
et al. 2019). Howe (2000) named this “inherent rarity,” which is related to
We found that only 22% of the species recorded in the the landscape configuration, and thus does not differ between
regional pool were unable to cross the matrix or to overcome native remnants and restoration patches in our study region.
the filters hindering germination and establishment in the under- Although zoochorous species in general successfully colo-
story of restoration forests. Biotic and abiotic filters have con- nized the restoration sites, 63% of the species not colonizing
strained seed dispersal as well as establishment of some the restoration forests were also dispersed by animals. Success
functional guilds, and that becomes clear when the functional of zoochorous species is not equally observed among animal
profile of unsuccessful colonizers (NC) is identified. Dispersal dispersers, the difference being related to distinct plant–animal
filters prevent species dispersed by gravity (autochory) or by ter- interactions (McAlpine et al. 2016). From our study, we infer
restrial mammals from crossing the matrix and arriving at the that dispersal by bats yields a higher colonization probability.
restoration islands, while slow-growing and rare species were The importance of bats for seed dispersal in tropical forests
likely filtered during the establishment phase, when these guilds has been already demonstrated (Galindo-Gonz
alez 1998;
can be eliminated by competition or predation. Galindo-Gonz
alez et al. 2000; Reid et al. 2015), and it was par-
The equally lower frequency of wind-dispersed and shade- ticularly successful for understory shrubs in our study. On the
intolerant individuals within both groups (successful and unsuc- other hand, dispersal by terrestrial mammals considerably
cessful colonizers) compared to the regional pool suggests that reduces the probability of reaching other isolated forest patches
their relative abundance in the regenerating community within or restoration sites. The absence of alternative dispersal agents
the restoration forests is similar to that in the native remnants. due to fragmentation and habitat loss has also been considered
These guilds have not been, therefore, filtered during the coloni- a barrier to tree dispersal. If the dispersal agents specific to the
zation process. Their lower relative abundance compared to unsuccessful colonizer species are not regionally extinct, their
their relative richness suggests filtering only within the restora- movement across the landscape has been likely limited by the
tion forests, with light competition being a candidate factor, agricultural matrix (Tambosi et al. 2013), which poses a serious
because most wind-dispersed trees are also shade intolerant. constraint for terrestrial mammals.
Their higher frequency in the regional pool would be likely Contrary to our expectations based upon a vast literature
inflated by the high proportion in which these guilds have been demonstrating significant relationships between diaspore
planted, as demonstrated by Vidal et al. (2020), and because dimensions (size and mass) and dispersal or establishment
they often occupy the overstory instead of the shaded under- (Osunkoya et al. 1992; Green & Juniper 2004; Moles
story, which was the focus of this study. et al. 2007; Lebrija-Trejos et al. 2016), none of the seed or dia-
Autochorous species depend primarily on gravity or on sec- spore traits analyzed showed to be a good predictor of coloni-
ondary dispersal to move over short distances within a forest zation success in our study. Most studies address one from
patch (Böhning-Gaese et al. 1999), and the dispersal barrier the two steps of colonization separately—seed dispersal or
among patches is particularly strong for this group of species plant establishment—while here we explored the outcome of
(Reid & Holl 2013). Competition combined with ontogenetic the two steps combined, therefore merging different biotic
issues are possible explanations for the relatively low proportion and abiotic filters. By merging the two steps, the positive rela-
of slow-growing individuals among the SC, compared to the tionship between seed mass and plant establishment (Moles
regional pool. Growing fast is advantageous when competition et al. 2007) was likely neutralized by the negative relationship
for resources, especially light, is a filter constraining plant estab- between seed mass and effectiveness of seed dispersal
lishment (Van Breugel et al. 2012). In addition, recruitment of (Howe & Smallwood 1982). Evidently, simply knowing about
slow-growing individuals may be delayed, as several years are seed dimensions does not allow us to predict if a tree species in
needed for a seedling from this guild to reach 50 cm in height, the surroundings will or will not successfully colonize the
which was the inclusion criterion used in our sampling. An even forest islands undergoing restoration.
longer time will be needed for the colonizing individuals to We found a certain degree of determinism in the colonization
become reproductive and begin to leave descendants within process in the study region, showing that the Field of Dreams
the forests undergoing restoration. It is, therefore, reasonable hypothesis is not always a myth (Hilderbrand et al. 2005). This

Restoration Ecology 7 of 10
Build it and most will come

determinism is related to both: (1) the vagility of the species, species, small populations that remain after habitat fragmenta-
which is dependent on the dispersal agent (D’Eon et al. 2002) tion or are re-introduced by restoration can quickly spread over
and degree of landscape fragmentation (Maina & Howe 2000), the landscape, thus facilitating their conservation. On the other
and (2) the environmental filters that guide the establishment hand, those species unable to cross the matrix or which fail to
of species under the canopy (Lebrija-Trejos et al. 2010), where establish in remnant or restoration forest islands face risks
temperature is more stable, vapor pressure is lower, light is less to their persistence. These species are condemned to gene ero-
available than outside, and competition for resources is intense. sion, or even local extinction if, for instance, the precise island
The portion of this determinism related to dispersal filters points refugium where they live is eliminated by land conversion or
to the high probability that less mobile species (autochorous or if the isolated populations are eliminated by a pest or disease,
short-distance dispersed by animals like ants with small ranges) among other factors (Bennett & Saunders 2010). For the unsuc-
will be confined to restored forest where they were planted or to cessful colonizers, persistence in the landscape requires great
the fragments where they occurred before fragmentation efforts such as seed collection, seedling production, enrichment
(Qian 2009), while those dispersed by flying vertebrates will planting, or even assisted migration, especially if the restoration
successfully colonize. The filtering mechanisms behind coloni- goal is to recover the full original biodiversity.
zation inevitably lead to a certain level of functional conver-
gence (Götzenberger et al. 2012) in the restoration forests. We
cannot disregard, however, that a multitude of unknown func- Acknowledgments
tions of individual species exist and that redundancy in one This research was financially supported by the Environment Secre-
function does not mean redundancy in any other function tariat of S~ao Paulo State/Global Environmental Facilities/World
(Naeem 1998). Despite this functional convergence in some Bank, the CNPq (grant nos. 302939/2009-1, 143423/2009-6 and
traits, there is no risk of biotic homogenization (Olden & Roo- 153844/2018-3), and FAPESP (grant no 2013/26470-3). The
ney 2006), because a previous study of these restoration forests authors thank J.M.D. Torezan, R. Machado, V. Pillar, H. Rocha,
(Suganuma & Durigan 2015) has shown their species diversity E. Camargo, I.G. Vieira, V.S. Almeida, and C. Machado, who pro-
continuously increasing toward reference levels. In addition, a vided information about functional traits. The authors thank
recent study with monospecific stands in the same region N. Guerin, S. Flake, M. Derhe, and two anonymous reviewers,
(Guerin et al. 2021) has shown that functional diversity is high who provided invaluable suggestions on the previous version of
and does not differ from remaining natural forests even if a this manuscript.
single species is planted.
We excluded from this study the exotic species recorded in
the forests being restored. However, our results also support pre-
LITERATURE CITED
dictions about the risk of invasion by these species. Assis
Aavik T, Helm A (2018) Restoration of plant species and genetic diversity
et al. (2013) recorded exotic species in all of 44 restoration pro-
depends on landscape-scale dispersal. Restoration Ecology 26:92–102
jects sampled in the same region of this study. We can predict Almeida-Neto M, Campassi F, Galetti M, Jordano P, Oliveira-Filho A (2008)
that, among the 93 non-native species recorded in these restora- Vertebrate dispersal syndromes along the Atlantic Forest: broad-scale
tion forests, those that are shade tolerant and dispersed by birds patterns and macroecological correlates. Global Ecology and Biogeogra-
or bats can potentially become invaders in the whole ecological phy 17:503–513
region. Our study allows us to also predict that the genetic diver- Amazonas NT, Viani RAG, Rego MGA, Camargo FF, Fujihara RT, Valsechi AO
sity of planted populations within the restoration forests tends to (2018) Soil macrofauna density and diversity across a chronosequence of
tropical forest restoration in southeastern Brazil. Brazilian Journal of
increase over time for the vagile species (the successful colo-
Biology 78:449–456
nizers in our study), as colonization by native genotypes from Assis GB, Suganuma MS, Melo AC, Durigan G (2013) Uso de espécies nativas
surrounding forest patches will likely occur. Indeed, an increase e ex
oticas na restauraç~ao de matas ciliares no estado de S~ao Paulo
in genetic variability over time within restoration forests has

(1957–2008). Revista Arvore 37:599–609
been already demonstrated for some tree populations (Zucchi Beltr
an LC, Howe HF (2020) The frailty of tropical restoration plantings.
et al. 2018), likely related to vagility. Restoration Ecology 28:16–21
Besides quantitatively demonstrating whether the Field of Bennett AF, Saunders DA (2010) Habitat fragmentation and landscape change.
Dreams hypothesis is valid for trees and shrubs in the seasonal Conservation Biology 93:1544–1550
Böhning-Gaese K, Gaese BH, Rabemanantsoa SB (1999) Importance of primary
Atlantic Forest, finding functional profiles for successful and
and secondary seed dispersal in the Malagasy tree. Commiphora
unsuccessful colonizers has the potential to inform restoration guillaumini Ecology 80:821–832
planning, monitoring, and adaptive management. Disentangling Brancalion PHS, Campoe O, Mendes JCT, Noel C, Moreira GG, Van Melis J,
the mechanisms behind successful or unsuccessful colonization Stape JL, Guillemot J (2019) Intensive silviculture enhances biomass
can provide relevant information to support ecological restora- accumulation and tree diversity recovery in tropical forest restoration.
tion. The way in which species move across isolated forest Ecological Applications 29:1–12
Burton JI, Mladenoff DJ, Clayton MK, Forrester JA (2011) The roles of environ-
patches in fragmented landscapes matters when strategies for
mental filtering and colonization in the fine-scale spatial patterning of
biodiversity conservation or restoration are designed (Reid & ground-layer plant communities in north temperate deciduous forests. Jour-
Holl 2013). For those species that easily cross the matrix and nal of Ecology 99:764–776
establish in other forest patches, conservation or restoration does Caiafa AN, Martins RF (2010) Forms of rarity of tree species in the southern
not require huge efforts of human intervention. For these Brazilian Atlantic rain forest. Biodiversity and Conservation 19:2597–2618

8 of 10 Restoration Ecology
 Build it and most will come

Carlucci MB, Brancalion PH, Rodrigues RR, Loyola R, Cianciaruso MV (2020) Jones HP, Jones PC, Barbier EB, Blackburn RC, Rey Benayas JM, Holl KD,
Functional traits and ecosystem services in ecological restoration. Restora- McCrackin M, Meli P, Montoya D, Mateos DM (2018) Restoration and
tion Ecology 28:1372 repair of Earth’s damaged ecosystems. Proceedings of the Royal Society
Carvalho PER (2003) Espécies arb
oreas Brasileiras. (1st ed.). Vol 1. B: Biological Sciences 285:20172577
Embrapa Informaç~ao Tecnol
ogica, Colombo, Brasília Colombo – PR: Kirmer A, Tischew S, Ozinga WA, Von Lampe M, Baasch A, Van
Embrapa Florestas Groenendael JM (2008) Importance of regional species pools and func-
Carvalho PER (2006) Espécies arb
oreas Brasileiras. (1st ed.). Vol 2. tional traits in colonization processes: predicting re-colonization after
Embrapa Informaç~ao Tecnol
ogica, Colombo, Brasília Colombo – PR: large-scale destruction of ecosystems. Journal of Applied Ecology 45:
Embrapa Florestas 1523–1530
Carvalho PER (2008) Espécies arb
oreas Brasileiras. (1st ed.). Vol 3. Laughlin DC (2014) Applying trait-based models to achieve functional targets for
Embrapa Informaç~ao Tecnol
ogica, Colombo, Brasília Colombo – PR: theory-driven ecological restoration. Ecology Letters 17:7717–7784
Embrapa Florestas Lebrija-Trejos E, Perez-Garcia A, Meave JA, Bongers FE, Poorter L (2010)
Catterall CP (2018) Fauna as passengers and drivers in vegetation restoration: a Functional traits and environmental filtering drive community assembly
synthesis of processes and evidence. Ecological Management and Restora- in a species-rich tropical system. Ecology 91:386–398
tion 19:54–62 Lebrija-Trejos E, Reich PB, Hern
andez A, Wright SJ (2016) Species with greater
Cross SL, Bateman PW, Cross AT (2019) Restoration goals: why are fauna still seed mass are more tolerant of conspecific neighbours: a key driver of early
overlooked in the process of recovering functioning ecosystems and what survival and future abundances in a tropical forest. Ecology Letters 19:
can be done about it? Ecological Management and Restoration 21:4–8 1071–1080
Crouzeilles R, Curran M, Ferreira MS, Lindenmayer DB, Grelle CEV, Rey Li L, Cadotte MW, Martínez-Garza C, de la Peña-Domene M, Du G (2018) Plant-
Benayas JM (2016) A global meta-analysis on the ecological drivers of ing accelerates restoration of tropical forest but assembly mechanisms
forest restoration success. Nature Communications 7:11666. https://doi. appear insensitive to initial composition. Journal of Applied Ecology 55:
org/10.1038/ncomms11666 986–996
D’Eon RG, Glenn SM, Parfitt I, Fortin M-J (2002) Landscape connectivity as a Litt AR, Provencher L, Tanner GW, Franz R (2001) Herpetofaunal responses to
function of scale and organism vagility in a real forested landscape. restoration treatments of longleaf pine sandhills in Florida. Restoration
Conservation Ecology 6:10 Ecology 9:462–474
Díaz-García JM, Pineda E, L
opez-Barrera F, Moreno CE (2017) Amphibian Lohbeck M, Poorter L, Paz H, Pla L, van Breugel M, Martínez-Ramos M, Bongers F
species and functional diversity as indicators of restoration success in trop- (2012) Functional diversity changes during tropical forest succession. Perspec-
ical montane forest. Biodiversity and Conservation 26:2569–2589 tives in Plant Ecology, Evolution and Systematics 14:89–96
Duarte LS, Carlucci MB, Fontana CS, Hartz S, Pillar V (2011) Plant diaspore

Lorenzi H (1998) Arvores Brasileiras: Manual de identificaç~ao e cultivo de
traits as indicators of mutualistic interactions in woody vegetation patches oreas do Brasil. (1st ed.). Vol 2. Editora Plantarum/LTDA,
plantas arb

developing into a grassland-forest mosaic. Community Ecology 12: Nova Odessa, Brasil
126–134

Lorenzi H (2002) Arvores Brasileiras: Manual de identificaç~ao e cultivo de
Durigan G, Siqueira MF, Franco GADC, Contieri W (2004) A flora arbusto-
oreas do Brasil. (4th ed.). Vol 1. Editora Plantarum/LTDA,
plantas arb

arbustiva do Médio Paranapanema: Base para a restauraç~ao de ecossiste-
Nova Odessa, Brasil
mas naturais. Pages 199–239. In: Vilas Bôas O, Durigan G (eds) Pesquisas


Lorenzi H (2009) Arvores Brasileiras: Manual de identificaç~ao e cultivo de plan-
em conservaç~ao e recuperaç~ao ambiental no Oeste Paulista: resultados da
oreas do Brasil. (1st ed.). Vol 3. Editora Plantarum/LTDA, Nova
tas arb

cooperaç~ao Brasil/Jap~ao. P
aginas e Letras, S~ao Paulo, Brasil
Odessa
Frick KM, Ritchie AL, Krauss SL (2014) Field of dreams: restitution of pollinator
Li L, Cadotte MW, Martínez-Garza C, de la Peña-Domene M, Du G (2017) Plant-
services in restored bird-pollinated plant populations. Restoration Ecology
ing accelerates restoration of tropical forest but assembly mechanisms
22:832–840
appear insensitive to initial composition. Journal of Applied Ecology 55:
Galindo-Gonz
alez J (1998) Dispersi
on de semillas por murciélagos: su importan-
986–996
cia en la conservaci
on y regeneraci
on del bosque tropical. Acta Zoo
Maina GG, Howe HF (2000) Inherent rarity in community restoration. Conserva-
Mexicana 73:57–74
tion Biology 14:1335–1340
Galindo-Gonz
alez J, Guevara S, Sosa VJ (2000) Bat- and bird-generated seed
Maron M, Hobbs RJ, Moilanen A, Matthews JW, Christie K, Gardner TA,
rains at isolated trees in pastures in a tropical rainforest. Conservation
Keith DA, Lindenmayer DB, McAlpine CA (2012) Faustian bargains?
Biology 14:1693–1703
Restoration realities in the context of biodiversity offset policies. Biological
Gonz
alez-Varo JP, Carvalho CS, Arroyo JM, Jordano P (2017) Unravelling seed
Conservation 155:141–148
dispersal through fragmented landscapes: frugivore species operate
unevenly as mobile links. Molecular Ecology 26:4309–4321 McAlpine C, Catterall CP, Nally RM, Lindenmayer D, Reid JL, Holl KD,
Götzenberger L, Bello F, Brathen KA, Davison J, Dubis A, Guisan A, Zobel M Thomson J (2016) Integrating plant-and animal-based perspectives for
(2012) Ecological assembly rules in plant communities–approaches, pat- more effective restoration of biodiversity. Frontiers in Ecology and the
terns and prospects. Biological Reviews 87:111–127 Environment 14:37–45
Green PT, Juniper PA (2004) Seed mass, seedling herbivory and the reserve Meloni F, Varanda EM (2015) Litter and soil arthopod colonization in reforested
effect in tropical rainforest seedlings. Functional Ecology 18:539–547 semi-deciduous seasonal Atlantic forests. Restoration Ecology 23:690–697
Guerin N, Mendes FBG, Cianciaruso MV, Suganuma MS, Durigan G (2021) Moles AT, Ackerly DD, Tweddle JC, Dickie JB, Smith R, Leishman MR,
Pure or mixed plantings equally enhance the recovery of the Atlantic forest. Mayfield MM, Pitman A, Wood JT, Westoby M (2007) Global patterns
Forest Ecology and Management 484:118932 in seed size. Global Ecology and Biogeography 16:109–116
Hilderbrand RH, Watts AC, Randle AM (2005) The myths of restoration ecology. Mori ES, Piña-Rodrigues FCM, de Freitas NP (2012) Sementes florestais: guia para
Ecology and Society 10:19 germinaç~ao de 100 espécies nativas. Instituto Refloresta, S~ao Paulo, Brasil
Howe HF, Smallwood J (1982) Ecology of seed dispersal. Annual Review of Myers JA, Harms KE (2011) Seed arrival and ecological filters interact to assem-
Ecology and Systematics 13:201–228 ble high-diversity plant communities. Ecology 92:676–686
IPEF (Instituto de Pesquisa e Estudos Florestais) (2016) Lista de Sementes de Naeem S (1998) Species redundancy and ecosystem reliability. Conservation

Biology 12:39–45
Arvores Nativas e de Essências Ex
oticas. http://www.ipef.br/sementes/
natexoticas.asp (accessed 11 Jul 2016) Nalon MA, Matsukuma CK, Pav~ao M & Ivanauskas (2020) Invent
ario Florestal
do Estado de S~ao Paulo, 2020, Instituto Florestal. https://www.

Restoration Ecology 9 of 10
Build it and most will come

infraestruturameioambiente.sp.gov.br/2020/08/novo-inventario-florestal- Suding KN (2011) Toward an era of restoration in ecology: successes, failures,

do-esp-aponta-crescimento-de-214-mil-hectares-de-vegetacao-nativa-no- and opportunities ahead. Annual Review of Ecology, Evolution, and Sys-
territorio-paulista/ (accessed 16 Feb 2021) tematics. 42:465–487
Noreika N, Kotiaho JS, Penttinen J, Punttila P, Vuori A, Pajunen T, Autio O, Suganuma MS, Durigan G (2015) Indicators of restoration success in riparian
Loukola OJ, Kotze DJ (2015) Rapid recovery of invertebrate communities tropical forests using multiple reference ecosystems. Restoration Ecology
after ecological restoration of boreal mires. Restoration Ecology 23:566–579 23:238–251
Olden JD, Rooney TP (2006) On defining and quantifying biotic homogeniza- Suganuma MS, Torezan JMD, Durigan G (2017) Environment and landscape
tion. Global Ecology and Biogeography 15:113–120 rather than planting design are the drivers of success in long-term res-
Oliveira CD, Oliveira IR, Suganuma MS, Durigan G (2019) Overstory trees in toration of riparian Atlantic forest. Applied Vegetation Science doi
excess: a threat to restoration success in Brazilian Atlantic forest. Forest 21:76
Ecology and Management 449:117453 Tambosi LR, Martensen A, Ribeiro MC, Metzger JP (2013) A framework to opti-
Osunkoya OO, Ash JE, Hopkins MS, Graham AW (1992) Factors affecting survival mize biodiversity restoration efforts based on habitat amount and landscape
of tree seedlings in North Queensland rainforests. Oecologia 91:569–578 connectivity. Restoration Ecology 22:169–177
Palmer MA, Ambrose RF, Poff NL (1997) Ecological theory and community res- Van Breugel M, Van Breugel P, Jansen PA, Martinez-Ramos M, Bongers F
toration ecology. Restoration Ecology 5:291–300 (2012) The relative importance of above-versus belowground competition
Piotto D, Craven D, Montagnini F, Ashton M, Oliver C, Thomas WW (2019) for tree growth during early succession of a tropical moist forest. Plant
Successional, spatial, and seasonal changes in seed rain in the Atlantic for- Ecology 213:25–34
est of southern Bahia, Brazil. PLoS One 14:1–15 Van Der Pijl L (1972) Principles of dispersal in higher plants. Spring-Verlag,
Qian H (2009) Beta diversity in relation to dispersal ability for vascular plants in New York
North America. Global Ecology and Biogeography 18:327–332 Vidal CY, Naves RP, Viani RA, Rodrigues RR (2020) Assessment of the nursery
Reid JL, Holl KD (2013) Arrival ≠ Survival. Restoration Ecology 21:153–155 species pool for restoring landscapes in southeastern Brazil. Restoration
Reid JL, Mendenhall CD, Rosales JA, Zahawi RA, Holl KD (2014) Landscape Ecology 28:427–434
context mediates avian habitat choice in tropical forest restoration. PLoS Wheeler CE, Omeja PA, Chapman CA, Glipin M, Tumwesigye C, Lewis SL
One 9:e90573 (2016) Carbon sequestration and biodiversity following 18 years of
Reid JL, Mendenhall CD, Zahawi RA, Holl KD (2015) Scale-dependent effects of active tropical forest restoration. Forest Ecology and Management
forest restoration on Neotropical fruit bats. Restoration Ecology 23:681–689 373:44–55
Ren Y, Lü Y, Fu B, Zhang K (2017) Biodiversity and ecosystem functional Wodika BR, Baer SG (2015) If we build it, will they colonize? A test of the field
enhancement by forest restoration: a meta-analysis in China. Land Degra- of dreams paradigm with soil macroinvertebrate communities. Applied Soil
dation and Development. 28:2062–2073 Ecology 91:80–89
Rey-Benayas JM, Newton AC, Diaz A, Bullock JM (2009) Enhancement of bio- Zucchi MI, Sujii PS, Mori GM, Viana JP, Grando C, Silvestre ED, et al. (2018)
diversity and ecosystem services by ecological restoration: a meta-analysis. Genetic diversity of reintroduced tree populations in restoration plantations
Science 325:1121–1124 of the Brazilian Atlantic Forest. Restoration Ecology 26:694–701
Sentelhas PC, Marin FR, Ferreira AS, S
a EJA (2003) Banco de dados clim
aticos
do Brasil. Embrapa Monitoramento por Satélite, Campinas, Brazil
Species Link (2016) Centro de Referência em Informaç~ao Ambiental – CRIA/ Supporting Information
Fundaç~ao de Amparo à Pesquisa do Estado de S~ao Paulo - FAPESP The following information may be found in the online version of this article:
http://splink.cria.org.br/ (accessed 11 Jul 2016)
Table S1. Location and characterization of the study sites.
Strickland MS, Callaham MA Jr, Gardiner ES, Stanturf JA, Leff JW, Fierer N,
Bradford MA (2017) Response of soil microbial community composition Table S2. Tree species sampled in 26 restoration sites and nine remnants of seasonal
and function to a bottomland forest restoration intensity gradient. Applied tropical forest, southeastern Brazil.
Soil Ecology 119:317–326 Table S3. Chi-square analysis of the frequency of functional traits, among species and
Sudduth EB, Hassett BA, Cada P, Bernhardt ES (2011) Testing the field of individuals, in three groups of species (two by two).
dreams hypothesis: functional responses to urbanization and restoration Table S4. Logistic regression analysis of probability of species from each dispersal
in stream ecosystems. Ecological Applications 21:1972–1988 syndrome (anemochory, zoochory by birds or bats).

Coordinating Editor: Mia Derhe Received: 2 December, 2020; First decision: 16 March, 2021; Revised: 18 May,
2021; Accepted: 22 August, 2021

10 of 10 Restoration Ecology